A dictionary for genomes

With sequence information in hand, the search for regulatory sites in promoters can be done by computers rather than cloning. But the primary tools for analysis, multiple-alignment algorithms, can only handle a small amount of sequence data. In the August 29 Proceedings of the National Academy of Sciences, Bussemaker et al. introduce an alternative algorithm that they dub 'MobyDick' (Proc Nat Acad Sci USA 2000, 97: 10096-10100). MobyDick treats DNA sequence as text in which allthewordshavebeenruntogether. It attempts to build a dictionary of 'words' by first finding over-represented pairs of letters. Letter frequency is used to determine the probability that the pairs exist thanks to chance, and this helps determine how larger fragments continue to be built. Bussemaker et al. test their algorithm on a space-less version of the first ten chapters of the novel Moby Dick, then attack a list of all of the upstream regions in the yeast genome. For yeast, approximately 500 dictionary entries fall above a plausible significance level, including 114 of the 443 experimentally confirmed sites, and good matches to approximately half of the motifs found in previous analyses of co-regulated genes, the cell cycle, and sporulation.